Wired telephone systems were originally designed to carry speech to enable voice conversations over long distances. More recently, public switched telephone systems have become a primary medium for transmitting not only voice, but also non-speech data, such as by use of facsimile machines that transmit image information over the telephone lines, or by modems that exchange digital data of various forms (text, binary executable files, image or video files) over these same phone lines.
Today, cellular and other wireless communication systems are in much greater use for purposes of both voice and data communication. Most cellular communication in use in the world today utilize either the GSM (including UMTS) or CDMA (IS-95 or CDMA2000) communication systems. These systems transmit voice data over a voice traffic channel using a modulated carrier wave. For example, 2G GSM uses GMSK modulation and IS-95 CDMA uses PSK modulation. Prior to modulating the voice data for wireless transmission, the voice input is run through a speech compression circuit such as a vocoder to compress the voice input into a smaller amount of data. This reduces the amount of voice data that needs to be transmitted via the wireless network, thereby permitting the use of a smaller bit rate and a greater number of users sharing the same communication system.
Various vocoder techniques have been proposed and used. The most common are various forms of linear predictive codings (LPC); for example, 2G GSM uses a RPE-LPC speech codec, while IS-95 CDMA uses a variable rate CELP codec. These predictive compression techniques are designed specifically for voice encoding and, as such, are designed to filter out noise and other non-speech components. As a result, the transmission of digital data (such as ASCII text, byte codes, binary files) can be problematic since the vocoder processing can corrupt the digital data, making it unrecoverable at the receiving end of the transmission. For example, the recently introduced Qualcomm™ 4G Vocoder is a CDMA2000 device that exhibits a time-varying, non-linear transfer function which, while acceptable for voice encoding, can impose significant distortion when attempting to transmit digital data via the vocoder.
The 4G vocoder uses the 3gpp2 standards-based EVRC-B codec having a full rate of 9.6 kbps. The codec also supports lower bit rates, including a 4.8 kbps half rate and a 1.2 kbps eighth rate. These lower rates are used when the vocoder determines that the full rate is not needed to adequately transmit the sound signals it receives. For example, background noise is typically transmitted at the one-eighth rate. The EVRC-B vocoder uses these different rates to achieve a target rate that can be controlled by the wireless carrier and, as a result, this overall encoding process can make it difficult to successfully send non-speech data through the vocoder.